---
title: " My portfolio"
author: "Jens Ruhof"
date: "2/20/2021"
output: 
  flexdashboard::flex_dashboard:
    storyboard: true
    social: menu
    source: embed
---

```{r setup, include=FALSE}
library(flexdashboard)
library(tidyverse)
library(spotifyr)
library(dplyr)
library(fmsb)
library(png)
library(plotly)
library(compmus)
library(gridExtra)
library(cowplot)
library(forcats)
library(knitr)
library(ggpubr)
library(tidymodels)
library(ggdendro)
library(heatmaply)
```

``` {r get_playlist, include=FALSE}
# Eminem overall
 Eminem <- get_playlist_audio_features("", "6E8sLG3bm1meTRj8n5l0ll")
mean_eminem <- Eminem %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```


``` {r get_playlist_lp, include=FALSE}
# Shady lp
shady_lp <- get_playlist_audio_features("", "35sidb3NwfwV5QYnri3yLv")
mean_shadylp <- shady_lp %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```


``` {r get_playlist_xv, include=FALSE}
shadyxv <- get_playlist_audio_features("", "3oJ5M47mmYILrqHyzuXctM")
mean_shadyxv <- shadyxv %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_rev, include=FALSE}
revival <- get_playlist_audio_features("", "7yqbChSYPODqONRuAw4noI")
mean_revival <- revival %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_ref, include=FALSE}
relapse_refill <- get_playlist_audio_features("", "1xyhz9gNz5jJIVTOA9v6Dh")
mean_relapse_refill <- relapse_refill %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_del, include=FALSE}
relapse_deluxe <- get_playlist_audio_features("", "1aQ1zZBcyBFDwsuOdm19cz")
mean_relapse_deluxe<- relapse_deluxe %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_rec, include=FALSE}
recovery <- get_playlist_audio_features("", "2FDhk6jh6iQUDCc7n73diZ")
mean_recovery<- recovery %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_ma, include=FALSE}
murdera <- get_playlist_audio_features("", "69NMqLx5Kszkwl0buUEoBA")
mean_murdera<- murdera %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_mb, include=FALSE}
murderb <- get_playlist_audio_features("", "5ZUIc7iAkEBBeEXzyj7eia")
mean_murderb<- murderb %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_show, include=FALSE}
eminemshow <- get_playlist_audio_features("", "5g2QlPXMIoH4j34xQiHE1J")
mean_eminemshow <- eminemshow %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_matherlp, include=FALSE}
matherlp<- get_playlist_audio_features("", "1ILmjmVb32khVNXG2Zw7kp")
mean_matherlp <- matherlp %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_matherlp2, include=FALSE}
matherlp2<- get_playlist_audio_features("", "2kmroX8obtAPEYaVPAaVHi")
mean_matherlp2 <- matherlp2 %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_kamikaze, include=FALSE}
kamikaze<- get_playlist_audio_features("", "6B5em5RI6brFjyK7eyos1h")
mean_kamikaze <- kamikaze %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r get_playlist_encore, include=FALSE}
encore<- get_playlist_audio_features("", "5VdzQrI0NliDdfwjQ0rPLN")
mean_encore <- encore %>%
  summarise(
    mean_danceability = mean(danceability),
    mean_energy = mean(energy),
    mean_speechiness = mean(speechiness),
    mean_acousticness = mean(acousticness),
    mean_instrumentalness = mean(instrumentalness),
    mean_liveness = mean(liveness),
    mean_valence = mean(valence)
    
  )
```

``` {r}
get_conf_mat <- function(fit) {
  outcome <- .get_tune_outcome_names(fit)
  fit %>% 
    collect_predictions() %>% 
    conf_mat(truth = outcome, estimate = .pred_class)
}

get_pr <- function(fit) {
  fit %>% 
    conf_mat_resampled() %>% 
    group_by(Prediction) %>% mutate(precision = Freq / sum(Freq)) %>% 
    group_by(Truth) %>% mutate(recall = Freq / sum(Freq)) %>% 
    ungroup() %>% filter(Prediction == Truth) %>% 
    select(class = Prediction, precision, recall)
}  
```




### Analysing Eminem

This corpus focusses on the rapper Eminem and how his style has changed through the years. The corpus uses the [Eminem complete playlist](https://open.spotify.com/playlist/6E8sLG3bm1meTRj8n5l0ll?si=ANkMvh1bS8CkKOzdnhOrYA), by Eminem. The playlist spans from Eminems first song "Just dont give a fuck" from his The Slim Shady LP in 1998 to the single "The adventures of moon man and slim shady" in 2020.

I chose this corpus, since Eminem is one of my favorite artists of all time. I find myself coming back to his songs on the regular and even went to his last tour in the Netherlands. Eminem has many singles and albums published through the years and all are surounded by some sort of controversy. Eminem even raps about the fact that some of his fans want the old slim shady back, but that he is not that person anymore. I want to check in what degree this statement holds when comparing his music through the years.

Eminem's total of eleven studio albums make a great music library to compare on various aspects. In these albums Eminem changed up his rap game by collabing with pop artists, such as Ed SHeeran and Rihanna and release more poppy songs. I want to see if this has happened more later on or if he did this before as well. Furthermore Eminem is known for his many words per second in songs such as "rap god" or "Godzilla". Words aren't used by the spotify API, but I expect to see a coorelation between other features and words per minute. One of these features is speechiness, since it states in the description that "values between 33% and 66% suggest both music and speech (such as rap)" [spotify](https://developer.spotify.com/documentation/web-api/reference/#object-audiofeaturesobject).

A possible weakness for the corpus might be the skits that Eminem puts in its albums. These aren't actually songs, but often help tell the story of the album. An example of this are the several Paul skits, wherein Eminem talks with his manager Paul.


***


Main focus points

- Least popular songs of Eminem according to Spotify.

- Most popular songs of Eminem according to Spotify.

- Interactive radarchart of Spotify measurements on Eminem's songs.

- Chromagram of the most popular song.

- SSM's based on chroma and timbre for two songs with differing style.

- Key vs Chordogram comparisson for two songs with differing style.

- Tempogram of a fast-paced Eminem song.

- Album predictions for 4 iconic albums.

- Results

![Eminem](/home/jensruhof/Documents/Computational_musicology/compmus2/eminem.jpg){width=100%}





### Least popular

``` {r} 
LEAST <- Eminem %>% 
  group_by(track.popularity) %>% 
  filter(track.popularity <= 35) %>%
  ggplot(aes(x =track.popularity , y = fct_reorder(track.name, track.popularity))) + 
  geom_col(aes(fill = track.album.name)) +
  labs(x= "Popularity") + 
  ggtitle("What are the least popular Eminem songs?", "Popularity ranking < 35") +
  theme(legend.position = "right",
        legend.text = element_text( size = 10),
        text = element_text(size = 10),
        axis.title.y = element_blank())

ggplotly(LEAST)
```


***


#### Least popular Eminem songs

- Eminem only has 17 'songs' that have a popularity of 35 or lower. Of these 17 'songs' only 2 are actually songs, namely "guts over fear" and "right for me".

- The remaining 15 'songs' are often skits or intro's to his albums. These skits are often found as the first song on an album and are meant to introduce the listener to the album. The Paul skits are an example of this. These skits often feature a phone conversation between Eminem and his music manager Paul Rosenberg, in which the album or their relation at that time is featured.

- The two songs that are part of Eminems least popular songs come from the same album
Shady XV, which is often regarded as his worst album. Thus seeing atleast one song from this album here isn't surprising.

### Most popular

``` {r}
most <- Eminem %>% 
  group_by(track.popularity) %>% 
  filter(track.popularity >= 75) %>%
  ggplot(aes(x =track.popularity , y = fct_reorder(track.name, track.popularity))) + 
  geom_col(aes(fill = track.album.name)) +
  labs(x= "Popularity") + 
  ggtitle("What are the most popular Eminem songs?", "Popularity ranking > 75") +
  theme(legend.position = "right",
        legend.text = element_text( size = 10),
        text = element_text(size = 10),
        axis.title.y = element_blank())

ggplotly(most)
```


***


#### Most popular Eminem songs
- Eminem has 11 'songs' that have a popularity of 75 or higher. 

- 9 of his 12 albums are featured in these graph. 

- 3 albums have 2 songs featured. These are The Eminem Show, Recovery and The marshall mather LP. These albums are seen as his better albums, which is in line with the popularity measurement. 

- Eminem presents the re-up (2006), ShadyXV(2014) and Revival(2017) aren't represented here.  The albums aren't considered his best and thus this isn't surprising.

- His most popular song (Godzilla) is from one of his most recent albums, indicating that Eminem still puts out great music.


### Radarplot

``` {r}
# test <- plot_ly(
#   type = "scatterpolar",
#   r= c(0.7129386,0.7213708,0.2985401,0.1928719,0.01342138,0.2951325,0.5395462),
#   theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
#   fill = "toself"
# )
# test <- test %>%
#   layout(
#     polar = list(
#       radialaxis = list(
#         visible = T,
#         range = c(0,1)
#       )
#     ),
#     showlegend = F
#   )
# 
# test
fig <- plot_ly(
    type = 'scatterpolar',
    fill = 'toself'
  ) 
fig <- fig %>%
  add_trace(
    r = c(0.7129386, 0.7213708, 0.2985401, 0.1928719, 0.01342138, 0.2951325, 0.5395462),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Overall'
  ) 
fig <- fig %>%
  add_trace(
    r = c(0.7283333, 0.6116333, 0.39715, 0.2735885, 0.05342238, 0.3095833, 0.6580333),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Shady LP'
  )
fig <- fig %>%
  add_trace(
    r = c(0.6994286, 0.7866071, 0.240125, 0.1382441, 0.0007273875, 0.2240429, 0.5306786),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Shady XV'
  )
fig <- fig %>%
  add_trace(
    r = c(0.68, 0.7187368, 0.2828474, 0.26261, 5.927158e-05, 0.3540158, 0.5248947),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Revival'
  )
fig <- fig %>%
  add_trace(
    r = c(0.6959655, 0.7850345, 0.3074414, 0.2102438, 3.313793e-07, 0.3116, 0.558),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Relapse Refill'
  )
fig <- fig %>%
  add_trace(
    r = c(0.7161364, 0.7728636, 0.3247182, 0.2411955, 4.368182e-07, 0.3461591, 0.551),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Relapse Deluxe'
  )
fig <- fig %>%
  add_trace(
    r = c(0.6685294, 0.9168824, 0.2466588, 0.1525941, 1.3e-07, 0.3044941, 0.5481765),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Recovery'
  )
fig <- fig %>%
  add_trace(
    r = c(0.714, 0.66645, 0.219065, 0.1089275, 0.0003467275, 0.31603, 0.431345),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Music to be murdered by side A'
  )
fig <- fig %>%
  add_trace(
    r = c(0.7418056, 0.6893056, 0.2627111, 0.1610169, 0.0004116275, 0.3235917, 0.5114167),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Music to be murdered by side B'
  )
fig <- fig %>%
  add_trace(
    r = c(0.72035, 0.604985, 0.28606, 0.1263365, 0.04794438, 0.17504, 0.504715),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'The Eminem Show'
  )
fig <- fig %>%
  add_trace(
    r = c(0.7386667, 0.6762222, 0.3176167, 0.1434128, 1.096333e-05, 0.3192111, 0.5775444),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'The Marshal Mathers LP'
  )
fig <- fig %>%
  add_trace(
    r = c(0.679125, 0.8578125, 0.3076563, 0.2053562, 0.00239375, 0.4185125, 0.521),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'The Marshal Mathers LP2'
  )
fig <- fig %>%
  add_trace(
    r = c(0.7768462, 0.6326154, 0.3834615, 0.2434692, 0, 0.2661923, 0.5315385),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Kamikaze'
  )
fig <- fig %>%
  add_trace(
    r = c(0.701087, 0.646, 0.2990478, 0.2437839, 0.008007407, 0.2993522, 0.5071261),
    theta = c("mean_danceability", "mean_energy", "mean_speechiness", "mean_acousticness","mean_instrumentalness", "mean_liveness", "mean_valence"),
    name = 'Encore'
  )
fig <- fig %>%
  layout(
    polar = list(
      radialaxis = list(
        visible = T,
        range = c(0,1)
      )
    )
  )

fig
```


***


#### Interactive radarchart containing all albums
- NOTE: This is an interactive radarchart. Meaning that you can toggle albums on and off to get a better view of the different albums. Recommended albums to select are:
his popular albums : Shady LP, Recovery, The mathers LP 1/2.
His not so popular items compared to a popular one: Shady LP, ShadyXV, revival.

- Looking at the popular albums they often spring out when comparing to the mean of all albums together a good example of this is recovery with a high mean_energy or the Shady LP with high speechiness.

- Looking at the least popular albums, they don't spring out in comparission with the mean of all albums together. This could indicate that these albums weren't daring enough for Eminems audience. 

- This isn't necesarrily true. When comparing the recent music to be murdered by albums with the overall means. Especially side B is inline with the overall mean, but still holds popular songs like 'GNAT' and is well received.


### Chroma of Godzilla
```{r}
godzilla<-
  get_tidy_audio_analysis("7FIWs0pqAYbP91WWM0vlTQ") %>%
  select(segments) %>%
  unnest(segments) %>%
  select(start, duration, pitches)

```
```{r}
godzilla %>%
  mutate(pitches = map(pitches, compmus_normalise, "euclidean")) %>%
  compmus_gather_chroma() %>% 
  ggplot(
    aes(
      x = start + duration / 2,
      width = duration,
      y = pitch_class,
      fill = value
    )
  ) +
  geom_tile() +
  labs(x = "Time (s)", y = NULL, fill = "Magnitude") +
  theme_minimal() +
  scale_fill_viridis_c() +
  ggtitle("Chroma sampling of Godzilla using euclidean") +
  theme(title = element_text(size = 9))

```

***


#### Godzilla (feat. Juice WRLD)

- This is a chromagram of the song Godzilla(feat. Juice WRLD). This song is the most popular song of Eminem and is all over the place. However, there are certain lines that are repeated throughout the process, namely: **C#|Db** , **D#|Eb** and **F#|Gb**. These lines show alternating magnitude intensity the most with a few exceptions. E.g. the **A#|Bb** line that sometimes has a magnitude of 1 aswell.

- Gozilla consists of Eminem and Juice WRLD rapping over an electronic umph that comes in steps of 3. These can be linked to the 3 main lines we see alternating in the cordogram. Another thing to note is that the magnitude lines are very thin. This is because Godzilla has a BPM of 166.

- Around the 160 second mark Eminem starts his staggering 10.65 syllables per second rap. This can be seen in the chromagram, since it is being introduced by a lower sound that is played for multiple seconds before he begins.
 


***


```{=html}

```


### SSM of Mockingbird and Tone deaf compared

```{r ssm info}
mockingbird <-
  get_tidy_audio_analysis("0KOYV2jgRZWLcLI94kiscs") %>% # Change URI.
  compmus_align(bars, segments) %>%                     # Change `bars`
  select(bars) %>%                                      #   in all three
  unnest(bars) %>%                                      #   of these lines.
  mutate(
    pitches =
      map(segments,
        compmus_summarise, pitches,
        method = "mean", norm = "manhattan"              # Change summary & norm.
      )
  ) %>%
  mutate(
    timbre =
      map(segments,
        compmus_summarise, timbre,
        method = "mean"              # Change summary & norm.
      )
  )
tone <-
  get_tidy_audio_analysis("1cEUi8QulMj1xgrPwwGC2p") %>% # Change URI.
  compmus_align(bars, segments) %>%                     # Change `bars`
  select(bars) %>%                                      #   in all three
  unnest(bars) %>%                                      #   of these lines.
  mutate(
    pitches =
      map(segments,
        compmus_summarise, pitches,
        method = "rms", norm = "manhattan"              # Change summary & norm.
      )
  ) %>%
  mutate(
    timbre =
      map(segments,
        compmus_summarise, timbre,
        method = "mean"              # Change summary & norm.
      )
  )
```

```{r ssm plots}
op <- mockingbird %>%
  compmus_self_similarity(pitches, "manhattan") %>% 
  ggplot(
    aes(
      x = xstart + xduration / 2,
      width = xduration,
      y = ystart + yduration / 2,
      height = yduration,
      fill = d
    )
  ) +
  geom_tile() +
  coord_fixed() +
  scale_x_continuous(breaks = seq(0, 266.424, by = 60)) +
  scale_y_continuous(breaks = seq(0, 266.424, by = 60)) +
  scale_fill_viridis_c(guide = "none") +
  theme_classic() +
  labs(x = "", y = "")
ot <- mockingbird %>%
  compmus_self_similarity(timbre, "euclidean") %>% 
  ggplot(
    aes(
      x = xstart + xduration / 2,
      width = xduration,
      y = ystart + yduration / 2,
      height = yduration,
      fill = d
    )
  ) +
  geom_tile() +
  coord_fixed() +
  scale_x_continuous(breaks = seq(0, 266.424, by = 60)) +
  scale_y_continuous(breaks = seq(0, 266.424, by = 60)) +
  scale_fill_viridis_c(guide = "none") +
  theme_classic() +
  labs(x = "", y = "")
rp <- tone %>%
  compmus_self_similarity(pitches, "aitchison") %>% 
  ggplot(
    aes(
      x = xstart + xduration / 2,
      width = xduration,
      y = ystart + yduration / 2,
      height = yduration,
      fill = d
    )
  ) +
  geom_tile() +
  coord_fixed() +
  scale_x_continuous(breaks = seq(0, 282.581, by = 60)) +
  scale_y_continuous(breaks = seq(0, 282.812, by = 60)) +
  scale_fill_viridis_c(guide = "none") +
  theme_classic() +
  labs(x = "", y = "")
rt <- tone %>%
  compmus_self_similarity(timbre, "euclidean") %>% 
  ggplot(
    aes(
      x = xstart + xduration / 2,
      width = xduration,
      y = ystart + yduration / 2,
      height = yduration,
      fill = d
    )
  ) +
  geom_tile() +
  coord_fixed() +
  scale_x_continuous(breaks = seq(0, 282.812, by = 60)) +
  scale_y_continuous(breaks = seq(0, 282.812, by = 60)) +
  scale_fill_viridis_c(guide = "none") +
  theme_classic() +
  labs(x = "", y = "")
```

```{r ssm grid}
grid.arrange (arrangeGrob(op, top=grid::textGrob("Mockingbird", hjust = 0.15), left="Chroma"),arrangeGrob(rp,top=grid::textGrob("Tone deaf", hjust = 0.21)),arrangeGrob(ot, left="Timbre"),rt,ncol=2, nrow=2, widths = c(2.2,2) ,heights = c(2.2,2), top=grid::textGrob("SSM, based on bars, time in seconds", hjust= 0.42))
```



***


#### Timbre and Chroma comparisson


- These Self Similarity matrices (SSM) are segmented by bars. The Mockingbird SSM's make use of the mean and manhattan distance, whereas the Tone deaf SSM for pitches uses rms and manahttan and for timbre the mean.

- The left side of the figure shows the chroma and timbre of Mockingbird. The chroma SSM shows several blocks, but aren't that clear. In the timbre based SSM these blocks become more obvious. Around the 90 seconds mark a clear switch can be seen. This is where the chorus of mockingbird starts and more instruments chime in. 

- The right side of the figure shows the chroma and timbre of tone deaf. The song tone deaf is very different from his usual style, the beat stays consistent, which can be seen by the diagonal line. In tone deaf Eminem remains at the same rithm the entire time and there aren't any other instruments used. This can be seen in the SSM's by the lack of patterns. 



***


```{=html}

```

```{=html}

```





### Key vs chorodgram analysis of Mockingbird vs Tone deaf

``` {r get_chors}
circshift <- function(v, n) {
  if (n == 0) v else c(tail(v, n), head(v, -n))
}

#      C     C#    D     Eb    E     F     F#    G     Ab    A     Bb    B
major_chord <-
  c(   1,    0,    0,    0,    1,    0,    0,    1,    0,    0,    0,    0)
minor_chord <-
  c(   1,    0,    0,    1,    0,    0,    0,    1,    0,    0,    0,    0)
seventh_chord <-
  c(   1,    0,    0,    0,    1,    0,    0,    1,    0,    0,    1,    0)

major_key <-
  c(6.35, 2.23, 3.48, 2.33, 4.38, 4.09, 2.52, 5.19, 2.39, 3.66, 2.29, 2.88)
minor_key <-
  c(6.33, 2.68, 3.52, 5.38, 2.60, 3.53, 2.54, 4.75, 3.98, 2.69, 3.34, 3.17)

chord_templates <-
  tribble(
    ~name, ~template,
    "Gb:7", circshift(seventh_chord, 6),
    "Gb:maj", circshift(major_chord, 6),
    "Bb:min", circshift(minor_chord, 10),
    "Db:maj", circshift(major_chord, 1),
    "F:min", circshift(minor_chord, 5),
    "Ab:7", circshift(seventh_chord, 8),
    "Ab:maj", circshift(major_chord, 8),
    "C:min", circshift(minor_chord, 0),
    "Eb:7", circshift(seventh_chord, 3),
    "Eb:maj", circshift(major_chord, 3),
    "G:min", circshift(minor_chord, 7),
    "Bb:7", circshift(seventh_chord, 10),
    "Bb:maj", circshift(major_chord, 10),
    "D:min", circshift(minor_chord, 2),
    "F:7", circshift(seventh_chord, 5),
    "F:maj", circshift(major_chord, 5),
    "A:min", circshift(minor_chord, 9),
    "C:7", circshift(seventh_chord, 0),
    "C:maj", circshift(major_chord, 0),
    "E:min", circshift(minor_chord, 4),
    "G:7", circshift(seventh_chord, 7),
    "G:maj", circshift(major_chord, 7),
    "B:min", circshift(minor_chord, 11),
    "D:7", circshift(seventh_chord, 2),
    "D:maj", circshift(major_chord, 2),
    "F#:min", circshift(minor_chord, 6),
    "A:7", circshift(seventh_chord, 9),
    "A:maj", circshift(major_chord, 9),
    "C#:min", circshift(minor_chord, 1),
    "E:7", circshift(seventh_chord, 4),
    "E:maj", circshift(major_chord, 4),
    "G#:min", circshift(minor_chord, 8),
    "B:7", circshift(seventh_chord, 11),
    "B:maj", circshift(major_chord, 11),
    "D#:min", circshift(minor_chord, 3)
  )

key_templates <-
  tribble(
    ~name, ~template,
    "Gb:maj", circshift(major_key, 6),
    "Bb:min", circshift(minor_key, 10),
    "Db:maj", circshift(major_key, 1),
    "F:min", circshift(minor_key, 5),
    "Ab:maj", circshift(major_key, 8),
    "C:min", circshift(minor_key, 0),
    "Eb:maj", circshift(major_key, 3),
    "G:min", circshift(minor_key, 7),
    "Bb:maj", circshift(major_key, 10),
    "D:min", circshift(minor_key, 2),
    "F:maj", circshift(major_key, 5),
    "A:min", circshift(minor_key, 9),
    "C:maj", circshift(major_key, 0),
    "E:min", circshift(minor_key, 4),
    "G:maj", circshift(major_key, 7),
    "B:min", circshift(minor_key, 11),
    "D:maj", circshift(major_key, 2),
    "F#:min", circshift(minor_key, 6),
    "A:maj", circshift(major_key, 9),
    "C#:min", circshift(minor_key, 1),
    "E:maj", circshift(major_key, 4),
    "G#:min", circshift(minor_key, 8),
    "B:maj", circshift(major_key, 11),
    "D#:min", circshift(minor_key, 3)
  )

```

``` {r get_info_chord}
mockingbird_chord <-
  get_tidy_audio_analysis("0KOYV2jgRZWLcLI94kiscs") %>%
  compmus_align(sections, segments) %>%
  select(sections) %>%
  unnest(sections) %>%
  mutate(
    pitches =
      map(segments,
        compmus_summarise, pitches,
        method = "mean", norm = "manhattan"
      )
  )

tone_chord <-
  get_tidy_audio_analysis("1cEUi8QulMj1xgrPwwGC2p") %>%
  compmus_align(sections, segments) %>%
  select(sections) %>%
  unnest(sections) %>%
  mutate(
    pitches =
      map(segments,
        compmus_summarise, pitches,
        method = "mean", norm = "manhattan"
      )
  )

mockingbird_key <-
  get_tidy_audio_analysis("0KOYV2jgRZWLcLI94kiscs") %>%
  compmus_align(sections, segments) %>%
  select(sections) %>%
  unnest(sections) %>%
  mutate(
    pitches =
      map(segments,
        compmus_summarise, pitches,
        method = "mean", norm = "manhattan"
      )
  )

tone_key <-
  get_tidy_audio_analysis("1cEUi8QulMj1xgrPwwGC2p") %>%
  compmus_align(sections, segments) %>%
  select(sections) %>%
  unnest(sections) %>%
  mutate(
    pitches =
      map(segments,
        compmus_summarise, pitches,
        method = "mean", norm = "manhattan"
      )
  )

```

``` {r plot_chord_key}
mbc <- mockingbird_chord %>% 
  compmus_match_pitch_template(
    chord_templates,         # Change to chord_templates if descired
    method = "euclidean",  # Try different distance metrics
    norm = "manhattan"     # Try different norms
  ) %>%
  ggplot(
    aes(x = start + duration / 2, width = duration, y = name, fill = d)
  ) +
  geom_tile() +
  # scale_x_discrete(breaks = seq(0, 266.424, by = 60)) +
  # scale_y_discrete(breaks = seq(0, 266.424, by = 60)) +
  scale_fill_viridis_c(guide = "none") +
  theme_minimal() +
  labs(x = "Time (s)", y = "") +
  theme(axis.text=element_text(size=5))

tc <- tone_chord %>% 
  compmus_match_pitch_template(
    chord_templates,         # Change to chord_templates if descired
    method = "euclidean",  # Try different distance metrics
    norm = "manhattan"     # Try different norms
  ) %>%
  ggplot(
    aes(x = start + duration / 2, width = duration, y = name, fill = d)
  ) +
  geom_tile() +
  scale_fill_viridis_c(guide = "none") +
  theme_minimal() +
  labs(x = "Time (s)", y = "") +
  theme(axis.text=element_text(size=5))


mbk <- mockingbird_key %>% 
  compmus_match_pitch_template(
    key_templates,         # Change to chord_templates if descired
    method = "euclidean",  # Try different distance metrics
    norm = "manhattan"     # Try different norms
  ) %>%
  ggplot(
    aes(x = start + duration / 2, width = duration, y = name, fill = d)
  ) +
  geom_tile() +
  # scale_x_discrete(breaks = seq(0, 266.424, by = 60)) +
  # scale_y_discrete(breaks = seq(0, 266.424, by = 60)) +
  scale_fill_viridis_c(guide = "none") +
  theme_minimal() +
  labs(x = "Time (s)", y = "") +
  theme(axis.text=element_text(size=5))

tk <- tone_key %>% 
  compmus_match_pitch_template(
    key_templates,         # Change to chord_templates if descired
    method = "euclidean",  # Try different distance metrics
    norm = "manhattan"     # Try different norms
  ) %>%
  ggplot(
    aes(x = start + duration / 2, width = duration, y = name, fill = d)
  ) +
  geom_tile() +
  scale_fill_viridis_c(guide = "none") +
  theme_minimal() +
  labs(x = "Time (s)", y = "") +
  theme(axis.text=element_text(size=5))


```

```{r chord_key_grid}
grid.arrange (arrangeGrob(mbc, top=grid::textGrob("Mockingbird", hjust = 0.15), left="Chord"),arrangeGrob(tc,top=grid::textGrob("Tone deaf", hjust = 0.21)),arrangeGrob(mbk, left="Key"),tk,ncol=2, nrow=2, widths = c(2.2,2) ,heights = c(2.2,2))
# ggarrange(mbc, tc, mbk, tk + rremove("x.text"),
#           ncol = 2, nrow = 2)
```


***

#### Explanation of Mockingbird and Tone deaf.

- The key and chordogram of mockingbird are all over the place. This is expected, since Mockingbird is known for its piano. The keygram shows the altering keys of the piano well over time. 

- Tone deaf howeber gives a very consistent result. The chordogram shows several blue lines which can be linked to the steady beat of tone dead. When looking at the keygram we see that many of these lighter lines are removed with a strong focus on the main two lines, which can be explained by the metronomish sound seen in Tone deaf.


***


```{=html}

```

```{=html}

```





### Tempograms
``` {r load tempos, cache = TRUE}
graveola <- get_tidy_audio_analysis("20oenBXlmwIfK0F3fQIjhM")

cyclic <- graveola %>%
  tempogram(window_size = 8, hop_size = 1, cyclic = TRUE) %>%
  ggplot(aes(x = time, y = bpm, fill = power)) +
  geom_raster() +
  scale_fill_viridis_c(guide = "none") +
  labs(x = "Time (s)", y = "Tempo (BPM)") +
  theme_classic()

non_cyclic <- graveola %>%
  tempogram(window_size = 8, hop_size = 1, cyclic = FALSE) %>%
  ggplot(aes(x = time, y = bpm, fill = power)) +
  geom_raster() +
  scale_fill_viridis_c(guide = "none") +
  labs(x = "Time (s)", y = "Tempo (BPM)") +
  theme_classic()

```

``` {r plotting, cache = TRUE}
grid.arrange (arrangeGrob(cyclic, top=grid::textGrob("cyclic", hjust = 0.15), left="Kim"),arrangeGrob(non_cyclic,top=grid::textGrob("non-cyclic", hjust = 0.21)),ncol=2, nrow=1, widths = c(2.2,2))
```


***

As you can see the cyclic vs non-cyclic tempo analysis has a very different estimation of BPM.
According to [get bpm](https://getsongbpm.com/song/kim/jq1ooB) Kim has a BPM of 142, which is more in line with the
cyclic estimation. The non-cyclic tempogram shows a tempo twice as high as the cyclic tempogram. A possible explanation for this is that most of the energy in a song occurs twice per beat and the cyclic tempogram does some correction, whereas the non-cyclic doesn't do this.

I chose for Kim, since it's an old song of Eminem and can be seen as an outlier, since the song is criticized as going lyrically to far.
Other sings i have tried showed an even more stabled rhythm, this steady beat could be explained by the use it has in the rap genre. It is much easier to rap in a consistent tempo. Eminem often raps 2 or 4 notes per beat.
Other songs i have tried where also more stable than this one, a possible explanation for this consistency could be the rap genre. Rap is often consistent in BPM, since the rhythm and beat is used to spit raps to, so clear lines where expected.


***


```{=html}

```


### Predicting a songs album
``` {r}
eminem_albums <-
  bind_rows(
    # shady_lp %>% mutate(playlist = "Shady_lp") %>% slice_head(n = 20),
    # shadyxv %>% mutate(playlist = "Shady XV") %>% slice_head(n = 20),
    # revival %>% mutate(playlist = "Revival") %>% slice_head(n = 20),
    # relapse_refill %>% mutate(playlist = "Relapse refill") %>% slice_head(n = 20),
    # relapse_deluxe %>% mutate(playlist = "Relapse Deluxe") %>% slice_head(n = 20),
    recovery %>% mutate(playlist = "Recovery") %>% slice_head(n = 20),
    # murdera %>% mutate(playlist = "Mustic to be murdered by A") %>% slice_head(n = 20),
    # murderb %>% mutate(playlist = "Mustic to be murdered by B") %>% slice_head(n = 20),
    eminemshow %>% mutate(playlist = "The Eminem Show") %>% slice_head(n = 20),
    # matherlp %>% mutate(playlist = "Mather LP") %>% slice_head(n = 20),
    matherlp2 %>% mutate(playlist = "Mather LP2") %>% slice_head(n = 20),
    kamikaze %>% mutate(playlist = "Kamikaze") %>% slice_head(n = 20)
    # encore %>% mutate(playlist = "Encore") %>% slice_head(n = 20)
    
  ) 
```

``` {r classify features}
eminem_features <-
  eminem_albums %>%  # For your portfolio, change this to the name of your corpus.
  add_audio_analysis() %>% 
  mutate(
    playlist = factor(playlist),
    segments = map2(segments, key, compmus_c_transpose),
    pitches =
      map(
        segments,
        compmus_summarise, pitches,
        method = "mean", norm = "manhattan"
      ),
    timbre =
      map(
        segments,
        compmus_summarise, timbre,
        method = "mean",
      )
  ) %>%
  mutate(pitches = map(pitches, compmus_normalise, "clr")) %>%
  mutate_at(vars(pitches, timbre), map, bind_rows) %>%
  unnest(cols = c(pitches, timbre))

```


``` {r classify recipe}
eminem_recipe <-
  recipe(
    playlist ~
      danceability +
      energy +
      loudness +
      speechiness +
      acousticness +
      instrumentalness +
      liveness +
      valence +
      tempo +
      duration +
      C + `C#|Db` + D + `D#|Eb` +
      E + `F` + `F#|Gb` + G +
      `G#|Ab` + A + `A#|Bb` + B +
      c01 + c02 + c03 + c04 + c05 + c06 +
      c07 + c08 + c09 + c10 + c11 + c12,
    data = eminem_features,          # Use the same name as the previous block.
  ) %>%
  step_center(all_predictors()) %>%
  step_scale(all_predictors())      # Converts to z-scores.
  # step_range(all_predictors())    # Sets range to [0, 1].
```

``` {r cross}
eminem_cv <- eminem_features %>% vfold_cv(5)
```


``` {r forest}
forest_model <-
  rand_forest() %>%
  set_mode("classification") %>% 
  set_engine("ranger", importance = "impurity")
eminem_forest <- 
  workflow() %>% 
  add_recipe(eminem_recipe) %>% 
  add_model(forest_model) %>% 
  fit_resamples(
    eminem_cv, 
    control = control_resamples(save_pred = TRUE)
  )

```

``` {r forest pred}
predi <-eminem_forest %>% get_pr()
# eminem_forest %>% get_conf_mat()

```

``` {r forest mosaic}
matrix <-eminem_forest %>% get_conf_mat() %>% autoplot(type = "mosaic")
```


``` {r}
important <-workflow() %>% 
  add_recipe(eminem_recipe) %>% 
  add_model(forest_model) %>% 
  fit(eminem_features) %>% 
  pluck("fit", "fit", "fit") %>%
  ranger::importance() %>% 
  enframe() %>% 
  mutate(name = fct_reorder(name, value)) %>% 
  ggplot(aes(name, value)) + 
  geom_col() + 
  coord_flip() +
  theme_minimal() +
  labs(x = NULL, y = "Importance")

```


``` {r}

lay <- rbind(c(1,1,1,1,2,2),
             c(1,1,1,1,2,2),
             c(1,1,1,1,2,2),
             c(1,1,1,1,2,2),
             c(NA),
             c(NA,NA,3,3,3,3))
grid.arrange(matrix, important, tableGrob(predi),layout_matrix = lay)

```

``` {r}
grid.arrange(matrix, important, tableGrob(predi) ,nrow=2, ncol =2)

```

*** 

- For classification I chose four well known EMinem albums that have mean outliers in comparisson with its other albums. This selection has been made based on the interactive radarplot. 
Recovery has a strong outlier in mean energy. The eminem Show has relatively high mean instrumentalness compared to the other albums. The marshal mathers LP has high mean energy and a high mean liveliness. Kamikaze has a high mean danceability. I expect my random forest to perform well based on these observations. It might struggle with recovery and the marshal amthers lp 2, since both of them score high on mean energy. For recipe I went with everything. 

- The results are in line with my expectations, namely that the eminem show and recovery would classify well. I expected this, since their values stood out. I also expected to see an overlap between the amthers lp2 and recovery, since the energy of the LP is very similar to recovery. However I expected a clearer division, because of the liveliness difference. Interesting enough the points on which i found the classification would go aren't the most important features for our forest.

- I believe this is due the outliers in every album that do not correspond with the mean values that can be seen in teh plot. The Eminem Show with the outlier in instrumentalness performs well and I think this is because of this outlier.




### Visualising the difficulties the random forest model had to face, by plotting the albums on the most important features.

``` {r}
test <- eminem_features %>%
  ggplot(aes(x = energy, y = loudness, colour = playlist, size = c01)) +
  geom_point(alpha = 0.65) +
  scale_color_viridis_d() +
  labs(
    x = "Energy",
    y = "Loudness",
    size = "Timbre component 1",
    colour = "Playlist"
  )
```
``` {r}
ggplotly(test)
```


***


This graph is plotted on 3 of the more distinct features the random forest classifies them, size being linked to the feature: timbre component 1. This graph is interactive. Meaning that albums can be turned off or on. This is usefull for the comparisson between two albums and understanding why the forest has difficulty distinguishing between them.

- on the previous page can be seen that the marshal mathers lp2 and recovery don't perform to well with regards to eachother. When looking at this graph you can see why.
- when looking at the eminem show and kamikaze it becomesobvious that the forest model performs better on these albums, since they have more outliers, which makes it easier for the model to distinguish them.

- Another interesting thing to see that Eminems albums are very similar when looking at features, since this is the clearest I could plot the graph. Using other features resulted in even more overlap.


### Results

Eminem has changed through the years, but his style not so much. In recent years many people asked for the real slim shady back. However he was never realy gone. It is true that some of his albums weren't in the unique Eminem style, but that isn't something that only happened in the 2010-2020s. Eminem has 3 albums that have flopped in comparisson with his succesfull albums , namely Eminem presents the Re-Up, ShadyXV and revival. However when looking at the release date of these albums they are spread out over the course of his career. 
Another thing to note is that critics didn't like Eminems first albums at first, but changed their opinion over time.  After the flopped albums ShadyXV and Revival many people said that Eminem has lost his touch, but with his three most recent albums (Kamikaze, Music to be murderd by, Music to be murdered by side B) he proved the haters wrong. 

What i have learned throughout this course is that people can say a lot about an artist, without substantial evidence to back it up. Eminems style has changed, but not necesarily for the worst. One could even argue that he needed the critique to get back on the top of his game. Every artist has tracks people dislike and that is okay. 

Furthermore at the start of this project I expected clearer differences between albums and songs. Even when looking at other corpi there weren't as many extremities as I had expected. his concludes me that looking at song features can give certain insights, but is not as complete or important as i had thought. The experience you feel when listening to music can't be seen in these measurements, since that differs per listener. 

The conclusion we draw is that every artist has its up and downs and that this is necesary in order to create their own style. Feature selection doesn't necesarily say everything, but it can give an estimation or guideline in what a listener sees as important or as good music. No artist or Listener should hold these features as the gold standard of their listening style, since I myself like one of Eminems Least popular songs (guts over fear) and Eminems most popular song (Godzilla) isn't my favourite.

These conclusions could benefit every music enjoyer, meaning that it doesn't matter what features a song consists of if you like it or not is up to you. As for the artist having music liked by everyone is your goal, but that doesn't mean you need to change your style or persona to get there.